Аннотация
One of the most important stages in the development of intonational text-to-speech systems is prosodic processing. The correct pause placement and pause duration play a large role for the perception of speech by human hearing. In this article, the authors propose a new way of modeling pause for the Kazakh language using a vector representation of words. Our group presents a model that uses the minimum amount of data obtained with the help of manual tagging. We believe that such an approach will be of interest for scientific groups working on the development of speech synthesis systems for poorly studied languages.